摘要
提出了一种基于动态主元分析的传感器故障检测方法。利用数据矩阵前t时刻和当前时刻的数据,建立多变量多时刻的自回归统计模型。计算主元数据矩阵,建立动态主元模型。以测量速度最慢的传感器的测量周期为统一采样周期,4个连续采样周期为一个诊断周期,建立动态三维测量矩阵,采用残差的平方预报误差的指数加权移动平均(Squared prediction error-Exponentially weighted moving average,SPE-EWMA)模型检测传感器故障。在只存在传感器故障的前提下,模拟发动机开车过程中几种典型的渐变性故障和突变性故障,实验结果表明,算法实时跟踪了各种检测指标的变化,准确检测出故障传感器。
A sensor falut detection method is presented based on dynamic principle component analysis. The self return statistic model is built by the former and current data of data matrix. The principle component data matrix is obtained to build dynamic principle component analysis (DPCA) model. The measurement period of the sensor with the slowest measurement speed is taken as the sampling rate; and four continuous sampling periods are selected as a diagnosis pe- riod to build dynamic and three-dimensional measurement matrix. The squared prediction er-ror-exponentially weighted moving average (SPE-EWMA) model is used to detect the sensor fault. Under the condition of the existence of the sensor fault, the gradual and sudden fault is simulated on the start process of aeroengine. Experimental results indicate that the algorithm can track the change of various indexes in real time and check out sensor faults.
出处
《数据采集与处理》
CSCD
北大核心
2008年第3期338-341,共4页
Journal of Data Acquisition and Processing
作者简介
李果(1980-),男,博士研究生,研究方向:飞行数据智能处理、多源信息融合,E—mail:airforce1980@126.com。
张鹏(1979-),男,博士研究生,研究方向:飞行数据智能处理。
李学仁(1963-),男,教授,研究方向:自动化检测技术。
魏瑞轩(1968-)男,教授,研究方向:先进飞行控制理论,无人机。
冀捐灶(1966-),男,副教授,研究方向:电力电子及其智能检测。